Solana team

updated 06 November 2018

710,000 tx/s with off-the-shelf hardware and no sharding. Scales with Moore's Law.

4.3
  • 4.2
    Site
  • 4.4
    Team
  • 4.5
    Idea
  • 4.6
    Tech
  • 4.0
    Media
  • 3.8
    Users

Copy this code to share our widget

Status: upcoming
Public sale: TBA-TBA
Cap: No info
Goal: No info
Find out more
    Anatoly Yakovenko
    CEO
    Greg Fitzgerald
    CTO
    Raj Gokal
    COO
    Eric Williams, PhD
    Data Science, Token Economics
    Stephen Akridge
    Engineering
    Alan Yu
    Partnerships, Biz Dev
    Michael Vines
    Principal Engineer
    Rob Walker
    Principal Engineer
    Pankaj Garg
    Senior Staff Engineer

Comments1

0

This project is still relatively new with much work to do.  According to the whitepaper investment in Solana will likely only be open to accredited investors so you'll need assets of over $1M or income greater than $200K/yr for two years to demonstrate that.

The goals of this project are to create a very high throughput blockchain that "scales with Moore's Law" ie. as CPUs get faster so will the blockchain.  Baseline predictions for throughput with a single node system are well in excess of 500k tps, actually more like 700k.  This is similar to the TPS of a modern commodities exchange like NASDAQ. It is unclear to me how this is affected as more nodes are added to the system and distributed across networks with higher latency, or what the upper bounds are on decentralizing by massively increasing the number of verifiers and validators.  The performance is given by having a single "leader" that uses a VDF - verifiable delay function - that is relatively slow to compute but much faster to verify which can be done in parallel across multiple GPU cores.  The leader interleaves transaction data with the output of the VDF which verifiably orders the transactions.  On top of a pool of verifiers, any one of which can become the leader, there is also a PoS based set of validator nodes which can lose their stake by slashing if they do their job incorrectly by consensus.  The validators provide finality to transactions within a few hundred milliseconds.  

Solana will also provide contracts written using extended Berkeley Packet Filter bytecode which is used the Linux kernel to filter network packets.   eBPF has a JIT and can be targeted by LLVM so theoretically contracts should be writable in any language LLVM supports including Rust.

The team building Solana is headed up by two former engineers from Qualcomm who did hardcore low-level stuff like operating system development and have a very pragmatic approach to building the software which is mostly written in Rust and NVIDIA's parallel processing language CUDA.  They have a lot of code online in Github which you can download and run standalone or join their public Testnet.  They have a dashboard monitoring the performance which you can check out now.

This is definitely a very interesting project to watch and even if it only has niche appeal it may gain support for applications that need very high throughput such as exchanges (commodities and advertising).